A CLOUD-BASED ARCHITECTURE PROPOSAL FOR REHABILITATION OF APHASIA PATIENTS

Rev. Rou. Sci. Techn. Électrotechn. et Énerg. Vol. 62, 3, pp. 332 337, Bucarest, 2017 A CLOUD-BASED ARCHITECTURE PROPOSAL FOR REHABILITATION OF APHASIA PATIENTS DORIN CÂRSTOIU 1, VIRGINIA ECATERINA OLTEAN 1, SANDA MARIA NICA 2, GABRIEL SPIRIDON 3 Key words: Aphasia, Cloud architecture, Interface, Database, Modular syste. Aphasia concerns neurological disorders of speech, with oderate to severe dysfunction of the echaniss involved in language expression and understanding, and influencing the ajor coponents of language behavior. Aphasia rehabilitation involves a speech language therapist and the principles of therapy regard the intensity of exercises execution, the use of ultiple fors of sensory stiuli and the increase of exercises difficulty. The therapy is expensive and has to be custoized to patients specificity, including their native language. This paper proposes, as a work in progress, a cloud-based architecture for aphasia rehabilitation of patients having Roanian as their native language. The plaltfor coprises an application database, an application logic and interfaces for patients and therapists. Accopanied by a virtual assistant, the patient has to solve exercises built by the therapist and, depending on the obtained scores, the treatent is adjusted and data can be stored in a statistic odule. A scoring procedure is proposed, as part of the application logic and back-up solutions for iniizing the estiated riksks during the project s lifecycle are forulated. 1. INTRODUCTION Aphasia represents a set of neurological language disorders resulting fro daage of the language areas of the brain, with ultiple causes, such as stroke or trauatic injury. Once the underlying cause has been treated, the priary treatent for aphasia is speech therapy that focuses on relearning and practicing language skills and using alternative or suppleentary counication ethods. According to the National Aphasia Association [1], aphasia affects about one in 250 people in the USA. Considering the iportance of counication in all aspects of life, getting help for a counication disorder is critical. A recent study fro the USA [2] showed that patients with aphasia experienced longer length of stays, greater orbidity, and greater ortality than did those without aphasia. The cost of edical assistance for patients with language disturbance are increased, both direct personnel, space, disposable ites, and indirect days of hospitalization, tie until return to work, change in work status, change in personal incoe, quality of life. As intensive edical treatent is ore efficient in reducing these associated costs, there is an increased interest for developing IT-supported solutions aied to assist and partially substitute the activity of the therapist [3]. It has to be ephasized that the use of IT-based applications constitutes a general trend also in other service sectors traditionally doinated by huanhuan interactions, like teaching [4]. Aphasia treatent depends on the langauge and cultural particularities of the patient [5] and, despite the existence of a nuber of software packages that are specially designed for rehabilitation of aphasia patients, ost of the are dedicated to native English patients. Presently, there are no cultural and linguistic-specific tools adapted for ost populations with regional languages, such as the Roanian population, this representing a doain where soe iproveent is required. This contribution presents a proposal for a cloud based architecture for apahsia rehabilitation of patients having Roanian as their native language. Aphasia types and rehabilitation principles are reviewed in Section 2, followed by a presentation of the principles of IT-supported aphasia rehabilitation tools, in Section 3. The proposed dedicated cloud-based platfor for aphasia rehabilitation, with an original logical odel, is finally presented in Section 4, followed by concluding rearks. 2. APHASIA REHABILITATION 2.1. TYPES OF APHASIA DISORDERS Aphasia concerns neurological disorders of speech, with oderate to severe dysfunction of the echaniss involved in language expression and understanding, and influencing the ajor coponents of language behavior (speech coprehension, spoken language repetition, object naing and speech fluency). After being diagnosed by a neurologist, the aphasia patient is subject to ore extensive tests, conducted by a speech-language pathologist, who perfors a coprehensive exaination of the person's counication abilities. There is no highly accurate classification for all aphasia types [6], but one can distinguish aong four ain types: Expressive aphasia the patient knows what he wants to say, but he has trouble expressing hiself. Receptive aphasia the patient hears the voice or sees the iages, but he is unable to ake sense of the words. Anoic aphasia the patient has trouble using the correct word for objects, places or events. Global aphasia the patient is not able to understand speech, speak, read or write. Aong the various coon syptos of aphasia, the following have to be taken into consideration, aong others: the inability to understand the language, the inability to pronounce (not due to uscle paralysis or weakness), 1 Politehnica University of Bucharest, Faculty of Control and Coputers, Spl. Independenței, 060042, Bucharest, Roania, E-ail:ecaterina.oltean@upb.ro 2 Colentina Clinical Hospital, Dept. of Neurology, Ștefan cel Mare Av., 19 21, Bucharest, Roania 3 IPA SA- Research Developent, Engineering and Manufacturing for Autoation Equipent and Systes, Calea Floreasca 169, 014459, Bucharest, Roania

2 A cloud-based proposal for rehabilitation of aphasia patients 333 the inability to speak spontaneously, the inability to for words, the inability to nae objects, poor enunciation, excessive creation and use of personal neologiss, the inability to repeat a phrase, persistent repetition of phrases, paraphasia (substituting letters, syllables or words), agraatis (inability to speak in a graatically correct fashion), dysprosody (alterations in inflexion, stress, and rhyth), using of incoplete sentences, the inability to read, the inability to write. In the general ters of inforation theory, one can regard these syptos as losses of efficiency of eaning of the transitted inforation, with the patient either sender or receiver [7]. 2.2. APHASIA REHABILITATION PRINCIPLES The rehabilitation principles rely on odular odels of the brain, which aid in the understanding of the brain as a coplex syste by decoposing it into structural odules or functional odules (scheas) and exploring the patterns of copetition and cooperation that yield the overall function [8]. Given the various sources, syptos and types of aphasia, there is no universal treatent for this language disorder. The traditionally used ethod of rehabilitation involves a speech language pathologist, because it focuses ostly on the ipairent. The goals of the rehabilitation plan are very specific to each patient, and they ust be agreed upon, before the rehabilitation progra is initiated. Several principles of therapy have been shown to iprove the outcoe of therapy [9, 10]: 1. Regardless of the type of therapy used, the outcoe is better if the intensity of therapy is increased. A given nuber of hours of therapy will yield a uch better outcoe if they are given in a few sessions over a few days rather than in any sessions over any days. 2. The effectiveness of aphasia therapy increases when therapists use ultiple fors of sensory stiuli. For instance, auditory stiuli in the for of usic, and visual stiuli in the for of pictures, drawings, are routinely used during aphasia therapy sessions. 3. Gradual increases in the difficulty of language exercises practiced during a given therapy session iproves the outcoe. 3. IT-SUPPORTED REHABILITATION TOOLS. STATE OF THE ART According to the sources in the literature [9 11], with all the advanceents in coputer technology today such as coputational speed, network bandwidth, streaing video, disk storage, recording technology, and copression software it is natural to see the coputerized treatent ethods ipleented in a variety of clinical settings. In this regard, IT-supported aphasia rehabilitation can be ipleented either as the only ethod of treatent, where the patient perfors the exercises alone and afterwards his or her perforance is reviewed by a clinician, or it can assist the clinician in showing the stiuli to the patient. There are two ain treatent approaches using IT-based rehabilitation tools: Substitute Skill Model, based on using a visual aid to iprove spoken language and Direct Treatent Model, based on using specific sets of exercises for treating different aphasia deficits. Nuerous studies, such as the ones reported in [12] and [13], were conducted to investigate the use of delivery therapy reotely via a coputer and Internet connection, under the direct supervision of a speech and language therapist. The results were encouraging, showing iproveents in language function outcoes and decreases in the tie required to be spent together with the therapist [11]. There are a nuber of existent software packages that are specially designed and aied for use in the rehabilitation of aphasia patients. Aong the are a freeware progra described in [14], the suite fro Bungalow Software [15], the Clicker-5 and CHAT progras fro Crick Software referred in [3] or Parrot Software's online progras [16]. All of these are intended for patients that have English as their native language. 4. A DEDICATED CLOUD-BASED REHABILITATION PLATFORM 4.1. OBJECTIVES The proposed dedicated cloud-based platfor is desired to be an innovative coputer-supported therapy and recovery solution, especially developed for the Roanian language and cultural particularities, as an aid in the rehabilitation of neurological patients with different types of aphasia syptos, using a series of fully-custoizable odules of exercises. The purpose is to substitute the presence of a speech-recovery therapist, while using the sae principles of speech and language therapy (SLT), the ethod that currently represents the best option for aphasia rehabilitation [17]. It is aied to be accessible to patients using obile devices, both in the clinical setting and also reotely, based on the new generation of cloud-coputing concepts. It is intended that the results be onitored and adjusted accordingly by clinicians, providing a odern, ore accessible and cost-efficient ethod of speech therapy. The application is desired to have a friendly user interface, aking it easy to use for both clinicians and patients and it is intended to be able to anage a potentially unliited nuber of patients, onitoring their evolution and allowing the clinician to adapt their therapy at any given oent. 4.2. THE MODULAR SYSTEM FOR EXERCISES To put in practice these goals, the application will provide a predefined set of treatent odules, established by experiented edical staff and based on scientific ethods. Each odule will be divided into different levels of difficulty. For each of these odules, a large nuber of exercises will be provided, designed for the Roanian language and aied at iproving the patients speech and

3 Dorin Cârstoiu et al. 334 (a) (b) Fig. 1 The odular syste for building exercises and odules, with objects recorded as visual sybols and written and spoken words: a) and the cloud based architecture b). language rehabilitation concerning: iage and sound recognition, iage association, iage and sound siilarities, word recognition, word association, verb and actions recognition, seantic classification of words, identification of objects, transcription of letters and words, solving arithetic operations, typing words, color identification. For each treatent odule, the clinician can add new exercises using objects fro a predefined set of objects, recorded as visual sybols and written and spoken word (Fig. 1a). Moreover, the clinician will also be able to define new treatent odules. 4.3. THE CLOUD-BASED ARCHITECTURE The application will be deployed in a cloud-based architecture, with a central server for the application, and both the therapists and the patients connecting to the server through internet enabled obile devices (for the patient) or portable coputers (for the therapists) (Fig.1b). The patients interface will be optiized for touch enabled obile devices, in order to provide ease of access even for those patients that are not coputer savvy. The application will also contain a statistics odule, allowing therapists to have an overall iage of the results produced by the treatents they have set up. The statistics odule will also provide a graphical and tabular evolution of patient scores for a certain type of exercise or odule. Based on these figures, an ultiate outcoe of the project will be establishing soe connections between certain aphasia syptos and diagnosis, and successful therapy schees. Thus, when a siilar patient is identified, the corresponding treatent and exercises will be used, leading to a ore rapid recovery. Each therapist will have access to the default odules and exercises, to the odules and exercises defined by hi and to odules and exercises shared with hi by other therapists. Also, each therapist will have access to his own patients but no to other therapists patients. 4.4. THE REHABILITATION PROGRAM The use of the application is preceded by a selection of a lot of patients with aphasia, based on criteria such as: C1. a clinical neurological exaination, C2. testing the ability to understand language, speak, read, write, C3. calculate and recognize usical signs (notes) atched for age, C4. sex ratio, C5. etiology of aphasia (stroke, tuor, head traua, deyelination, infectious disease, deentia) and C6. subtypes of aphasia (expressive, receptive, anoic and global). Each patient in a lot will benefit fro a specific edical treatent in order to organize a scientific, coputerized, speech rehabilitation progra sustained on visual, auditory and interactive skills (repeating, answering, reading, writing, and calculating). When establishing the treatent schee for a specific patient, the therapists can visually cobine (drag and drop) exercises fro different odules and difficulty degree, or fro the custo exercises created by the therapists theselves, according to each patient s specific needs. The scores obtained by the patient during his treatent, his overall progress and a set of predefined paraeters, like total tie needed to solve an exercise, are onitored and stored in the database of the application. Thus, at any tie, the therapist can view the results of the conducted therapy and check on the progress and iproveents ade by the patient. Based on these previous results, it will be easier to decide upon the continuation of the treatent. Throughout the use of the application, a virtual assistant, encouraging hi in doing the exercises, will accopany the patient. A large database of patients with different fors of aphasia will be constructed, which will have a great ipact on the study and rehabilitation of aphasia. 4.5. THE LOGICAL MODEL The logical odel of the rehabilitation progra has two coponents: a structural part, referring a foralization of data and inforation and the logical procedure anaging the interaction of the patient with a specific treatent odule.

4 A cloud-based proposal for rehabilitation of aphasia patients 335 4.5.1. Foralization of data and inforation The foralization of data and inforation regards the lots of patients and the odules containing exercises. Given N a set of patients P = { p i } i= 1, each patient p i is evaluated according to the selection criteria C1:6 in section 4.4 and placed, by the therapist, in one of the subsets coposing the partition P = k=1: K Pk. Obviously, patients fro distinct subsets ay share soe evaluation results following the selection. The structure of the lots of patients is the decision of the therapist, despite the fact that a sei-autoatic selection procedure can be built as support to this purpose. Consider an indexed set of basic exercises = }. J B { b j j = 1 Each b j is created by the therapist (Fig. 1a) and stored in the database for configuring different odules. For each M k partition subset P k, a set of odules Modk = { od } = 1 is active at a given oent, with the therapist responsible for controlling the structure dynaics. A odule od = ex : j I, n = 1: N } is an { jn ordered set of therapeutic exercises ex jn, where I { 1: J} is a subset of indexes of corresponding basic exercises and n is the ordering index in the odule: ex jn precedes ex j( n+ 1) and is preceded by ex j( n 1). A therapeutic exercise, as coponent of the odule od Mod k, is defined as a tuple: ex jn = ( b j, N j, Tj, diff j ), where b j B is the chosen basic exercise, N j is the axial nuber of repetitions of b j in a tie interval of length less or equal to the value T j, estiated by the therapist to be necessary to reach success and diff j is the associated degree of difficulty established by the therapist. Observe that N j and T j together odel the recoended intensity of the exercise. 4.5.2. A procedure for scoring a therapeutic exercise The therapeutic exercises are repeated by the patient the iposed nuber of ties, and each exercise perforance is evaluated, in order to copute the score associated to the dyad patient-exercise. Also, the selected exercise has to be repeated in a tie interval no longer than the specified liit. For a selected therapeutic exercise ex jn, consider the variable cont { 0 : N j}, defining the counter of exercise repetitions. The patient s response to the exercise, response_ b j, during the current cycle repetition, can be correct, incorrect, or it ay not be sent in tie. If patient s response is received in tie, then it is copared to the correct solution by the evaluation function eval :{ response _ b j } {0,1}, with 0 for failure and 1 for success. The history of patient s response evaluations is updated, at each cycle, with the current evaluation and stored in the variable value. The coputed final score is SCOR = value diff j N j and the variable tier is the current output of a clock easuring the tie for the allowed axial interval T j. The interaction patient-exercise begins with the signal start_ b sent by the syste to the patient. If an exercise j cycle is carried out and response_ b j is received, then the current perforance is evaluated and a new cycle starts. The therapeutic exercise ends norally if the total nuber of cycles N is reached. If the allocated axial j tie interval T j runs out and the exercise is not ended, then the execution is aborted and the partial results are stored (Fig. 2). jn Procedure score( ex ) Input: b j, T j, diff j, N j. Output: cont, tier, SCOR. begin cont =0, value = 0 start tier while ( N j - cont)>0 cont = cont + 1 %to the patient send start_ b j wait_ response_ b j %fro the patient if receive response_ b j value = value + eval( response _ b j ) %fro the clock elseif tier == T j break end end stop tier SCOR = value diff j N j disp( cont, tier, SCOR ) end After finishing all the therapeutic exercises in the odule, ex jn od, n = 1: N, the patient is reevaluated by the therapist who interprets all the output results score( ex jn ), and according to this evaluation, the therapist eventually reconfigures a new odule, or launches an already elaborated odule of therapeutic exercises to be executed. Alternatively, if the ean score is too low, the therapy is interpreted as inefficient and a decision is taken accordingly by the clinician. 4.6. THE PROJECT STEPS AND SOME ASPECTS REGARDING THE RISKS ASSESSMENT The design of the cloud-based architecture and the ipleentation of the rehabilitation progra will be developed in three ain steps: (1) the specification of the user requireents and of the syste architecture and, in parallel, the construction of the lots of patients; (2) the design and developent of the prototype, its integration and

5 Dorin Cârstoiu et al. 336 deployent, iplying also the accepted interface prototype and the functional testing of the alpha version; (3) the evaluation of the results of the coputerized platfor on the selected lots of patients, disseination and exploitation of the application. Roanian language has a coplex vocabulary and graar and the rehabilitation has to be ade in the native language of the patients. It is worth entioning that the ere translation in Roanian of existing software and/or tests is not a viable Fig. 2 A siplified view of the inforation exchanged by the procedure score with the users. During the entire project lifecycle, a constant assessent and evaluation of risks is required, iplying continuous onitoring and counication between partners and stakeholders. Using, as tool, the classic SWOT (strengths, weaknesses, opportunities and threats) analysis, several possible project risks have been identified, with their specific risks levels and associated back-up solutions, aied to iniize those risks levels (Table 1). Risk / level Table 1 Estiated risks and associated back-up solutions the specifications for the odule requireents can be inadequate / low the software partners can isunderstand the edical requireents / low the patients ay present an inability to use the therapeutically devices / ediu the obile devices ay be subject of generational change and obsolescence / low delays in testing and developent ay appear / low lack of interest to coercialize the syste / ediu Back-up solution aied to iniize the risk level the clinical and inforatics skills and experience of the participants will prove adequate in preventing it the active participation of the software partners in user requireents definition perforing the patient s interaction through an intuitive interface based on touch-input the fact that the new technologies are friendlier to the patient user involving the ain stakeholders early in the project, inforing the about the project s steps and advantages involving the ain stakeholders early in the project, inforing the about the project s steps and advantages 5. CONCLUSIONS Although other countries developed coputer progras for the recovery of the aphasic patient, these progras cannot be used in Roania, due to language barrier, as the option in this case, due to the structural linguistic differences in word phonology (different phonees and cobinations of phonees) and word fors (different lengths, syllables) that require distinct cognitive processing and should be adapted to the native language of the patients. The proposed project of a cloud-based aphasia rehabilitation architecture uses a Direct Treatent Model to ipleent the principles of aphasia rehabilitation intensity increase, ultiple fors of stiuli and gradually increase in difficulty. The structure coprises an application database, application logic and interfaces with the therapist and the patients, respectively. The treatent will be custoized for lots of patients and for each patient in a lot, and will consist of intensive repetition of exercises fro specially built odules. The odules can be reconfigured according to the particular treatent in progress. Each exercise in a odule is characterized by intensity and by the level of difficulty. A scoring procedure for the evaluation of patients response to exercises is proposed, as part of the application logic. As part of the risk assessent and counication strategy during the project s lifecycle, several back-up solutions aied to iniize the estiated risks levels have been forulated. Future research regards the ipleentation of the application. Received on January 9, 2017 REFERENCES 1. *** National Aphasia Association, http://www.aphasia.org/aphasiadefinitions/ 2. C. Ellis, A.N. Sipson, H. Bonilha, P.D. Mauldin, K.N. Sipson, The one-year attributable cost of post stroke aphasia, Stroke, 43, 5, pp. 1429 1431 (2012). 3. Joan L. Green, Coputer Progras for Expressive Aphasia, 2008, http://www.speechpathology.co/ask-the-experts/coputer-prograsfor-expressive-aphasia-996 4. Anca Daniela Ioniță, Adriana Olteanu, Support students experiental work in electronical engineering with visual odeling, Rev. Rou. Sci. Techn. Électrotechn. et Énerg., 59, 1, pp. 107 116 (2014).

6 A cloud-based proposal for rehabilitation of aphasia patients 337 5. G. Vigliocco, S. Kita, Language-Specific Effects of Meaning, Sound and Syntax: Iplications for Models of Lexical Retrieval, Production Language and Cognitive Processes, 21, pp. 790 816 (2006). 6. B. Kolb, I.Q. Whishaw, Fundaentals of Huan Neuropsychology, 6 th ed., 2007, New York, N.Y Worth Publishers. 7. R. Dobrescu, D. Merezeanu, Fro inforation to knowledge transission of eaning, Rev. Rou. Sci. Techn Électrotechn. et Énerg., in press (2017). 8. R.C. Berwick, A.D. Friederici, N. Chosky, J.J. Bolhuis, J.J., Evolution, brain, and the nature of language. Feature Review, Trends in Cognitive Sciences, 17, 2, 89 98 (2013). 9. Aphasia Rehabilitation Best Practice Stateents, Coprehensive suppleent to the Australian Aphasia Rehabilitation Pathway, www.aphasiapathway.co.au 10. J. Stark, N. Martin, R. Fink (Eds.), Aphasia Therapy Workshop. Current approaches to aphasia therapy. Principles and applications, A special issue of Aphasiology, 19, 10/11, Taylor & Francis Group, Hove, UK, Psychology Press, 2005. 11. J. Mortley, J. Wade, P. Enderby, A. Hughes, Effectiveness of Coputerised Rehabilitation for Long-Ter Aphasia: A Case Series Study, British Journal of General Practice, 54, pp. 856 857 (2004). 12. S.J. Doesborgh, M.W. van de Sandt-Koenderan, D.W. Dippel, F. van Harskap, P.J. Koudstaal, E.G. Visch-Brink, Effects of Seantic Treatent on nverbal Counication and Linguistic Processing in Aphasia after Stroke: A Randoized Controlled Trial, Stroke, 35, 1, pp. 141 146 (2004). 13. W.M.E. van de Sandt-Koenderan, J. Wiegers, Sandra M. Wielaert, H.J. Duivenvoorden, G.M. Ribbers, High-tech AAC and severe aphasia: Candidacy for TouchSpeak (TS), Aphasiology, 21, 5, pp. 459 474 (2007). 14. *** Power AFA AFASIA, software.inforer. http://powerafaafasia.software.inforer.co/ 15. *** Bungalow Software. Speech & Language Therapy for Aphasia. http://bungalowsoftware.co/aphasia/ 16. *** ParrotSoftware, Effective Treatent for Aphasia and Brain Injury. http://www.parrotsoftware.co/ 17. Marian C. Brady, Helen Kelly, J. Godwin, Pa Enderby, Pauline Capbell, Speech and language therapy for aphasia following stroke, Cochrane Database of Systeatic Reviews 2016, Issue 6, Art. No.: CD000425, John Wiley & Sons, Ltd, http://onlinelibrary. wiley.co